Deep research you can follow — claim by claim, source by source.
24 reports · 6,347 sources read · 452 receipts attached
How an AI Agent Should Decide Which Decisions Deserve Credit or Blame
When an AI agent succeeds or fails, how should it decide which earlier decisions deserve credit or blame—and what should change as a result?
A final score describes the outcome of an entire trajectory, while an update must act on some particular part of the system. That gap exists whether the run succeeds or fails.
Treat the outcome as evidence, not an explanation. Reconstruct the decision and what the agent knew at the time; compare only alternatives it could actually have chosen; then replay or approximate those alternatives and estimate how much they change the outcome, with uncertainty left visible. Separate plan errors from execution errors, and do not force delayed or interacting effects into a precise blame score when the evidence cannot support one. Turn the diagnosis into the smallest reversible change, test it against the incumbent, and abstain or collect more data when replay fidelity, coverage, or downside risk is inadequate.
We read 216 sources and cited 14. Every citation ships with a receipt — open one:
What must be remembered in order to reconstruct the decision actually made at the time?
The OpenTelemetry GenAI attributes listed on this page are marked as Deprecated and Moved to the OpenTelemetry GenAI semantic conventions repository. The schema provides attributes for identifying the agent and conversation context, including gen_ai.agent.id, gen_ai.agent.name, gen_ai.agent.version, and gen_ai.conversation.id. The schema provides attributes for capturing model request settings and response outcomes, including gen_ai.request.temperature, gen_ai.request.top_p, gen_ai.request.top_k
What must be remembered in order to reconstruct the decision actually made at the time?
Agent-learning work likewise formalizes a trajectory as an instruction followed by action–observation history, step, transition, and outcome.
Given an expert trajectory dataset D = {(u, e)}, where each trajectory e = (u, a_1, o_1, ..., a_m, o_m), u represents the initial task instruction, a_t denotes the action (including its rationale) at step t, o_t is the corresponding observation.
What must be remembered in order to reconstruct the decision actually made at the time?
Visible reasoning is not a complete causal record either: across six prompted-hint settings, one study found that reasoning traces often disclosed a used hint in fewer than 20% of affected examples.
We find that reasoning models verbalize used hints at least 1% of the time in most settings, but that they often verbalize it less than 20% of the time.
Every report is tagged by the ground it covers; each tag is a standing thread.